hermes-cloudflare

Cloudflare Browser Rendering plugin for hermes-agent — crawl, scrape, and extract content from web pages using Cloudflare's headless browser API.

ℹ️ Cloudflare has renamed Browser Rendering to Browser Run — same API, endpoints unchanged.

Tools

Tool	Description
cf_crawl	Async website crawling with depth/limit controls, format selection (HTML/Markdown/JSON), include/exclude patterns, sitemap discovery
cf_scrape	CSS-selector-based element extraction (text, HTML, attributes, dimensions)
cf_markdown	Convert pages to clean Markdown (handles JS-heavy sites)
cf_json_extract	AI-powered structured data extraction (Workers AI / Llama 3.3 70B) with prompt + JSON schema
cf_links	Link discovery with visible-only and external domain filtering
cf_content	Fully rendered HTML after JavaScript execution
cf_screenshot	Page screenshots (full-page, element, viewport control, PNG/JPEG/WebP)
cf_pdf	Page-to-PDF with headers/footers, margins, scale
cf_snapshot	Multiple formats (HTML, screenshot, Markdown, a11y tree) in a single request
cf_accessibility_tree	Accessibility tree of a page (roles, names, values) after JS execution
cf_ai_chat	Workers AI text generation via REST (Llama models; needs "Workers AI - Read" token permission)
cf_dns	DNS-over-HTTPS record lookups (A, AAAA, MX, TXT, …) — no credentials needed

Saving binary output

Screenshots and PDFs can exceed the inline response limit. Instead of returning base64, save them directly:

• output_path — write the decoded file locally (supported by cf_screenshot, cf_pdf, cf_snapshot)
• r2_bucket + r2_key — upload to an R2 bucket (requires "Workers R2 Storage - Edit" token permission)

cf_pdf also accepts raw html instead of a url for generating PDFs from custom HTML/CSS.

Reliability

Read endpoints are cached for 5 minutes (renders are billed per browser-second), and requests are automatically retried with backoff on rate limits (429) and transient server errors (5xx).

Installation

Quick install (hermes-agent v0.3.0+)

curl -sSL https://raw.githubusercontent.com/raulvidis/hermes-cloudflare/main/install.sh | bash

Manual install

git clone https://github.com/raulvidis/hermes-cloudflare.git
mkdir -p ~/.hermes/plugins
cp -r hermes-cloudflare/hermes-cloudflare-plugin ~/.hermes/plugins/hermes-cloudflare
pip install httpx

Configuration

Set these environment variables (in your shell profile or .env):

export CLOUDFLARE_API_TOKEN="your-api-token"
export CLOUDFLARE_ACCOUNT_ID="your-account-id"

Getting your credentials

1. Go to Cloudflare Dashboard
2. Create a token with Browser Rendering - Edit permission (add Workers AI - Read for cf_ai_chat and Workers R2 Storage - Edit for R2 uploads)
3. Copy your Account ID from the dashboard sidebar

Usage examples

Once installed, hermes-agent can use the tools directly:

Crawl a website:

Crawl https://docs.example.com and return the content as Markdown

Scrape specific elements:

Scrape all h1 tags and .price elements from https://shop.example.com

Extract structured data with AI:

Extract all product names, prices, and ratings from https://store.example.com as JSON

Get page as Markdown:

Convert https://blog.example.com/post to Markdown

Take a screenshot:

Take a full-page screenshot of https://example.com

Capture multiple formats at once:

Get the Markdown, screenshot, and accessibility tree of https://example.com in one request

Inspect page structure:

Show me the accessibility tree of https://app.example.com so I can see the interactive elements

Save a PDF of a page:

Render https://example.com/report as a PDF and save it to /tmp/report.pdf

Resolve a DNS record:

What are the MX records for example.com?

Requirements

• hermes-agent v0.3.0+
• Python 3.10+
• httpx (installed automatically by install.sh)
• Cloudflare account with Browser Rendering enabled (available on Free and Paid plans)

License

MIT